How it works

One agent that reasons, calls the tool, and renders the result as a card.

Fahd isn't a feature list — it's one agent on the Claude Agent SDK that understands the student's or teacher's intent, picks the right tool from 31 capabilities, fetches data from your platform, then renders it as a readable card — never raw text or JSON.

The agent loop

Four steps between the question and the card.

1

Understands intent

Reads the question in the context of the course, the user's role, language, and grammatical gender — not in isolation.

2

Picks the tool

Decides which of 31 capabilities serves the intent — fetch assignments, search content, detect at-risk learners…

3

Fetches from your platform

Calls the tool on your platform (Blackboard / Moodle / edX) within role permissions and platform limits.

4

Renders as a card

Summarizes the result in a tidy card; large composites condense with an "open in full screen" affordance.

Result cards, not raw text

Every tool call renders like this.

Live examples from the PHYS 101 course — built with Fahd's own design-system components.

The registry

31 tools. Capabilities, not buttons.

Fahd loads its tool registry at boot; available capability varies by platform. What your platform doesn't support shows as a dimmed capability — not a broken button.

Compare capabilities across platforms

get_assignmentsget_gradesget_calendarsearch_contentget_content_treeget_discussionsget_filesget_groupsget_attendanceget_rubricdraft_announcementget_masteryclass_overviewwhat_if_gradeget_submissionslist_modulesget_announcementssend_message

A sample of the registry. send_message is unsupported on edX — Fahd offers an alternative.

Graceful degradation

When a capability is missing, Fahd offers an alternative — it doesn't break.

No generic errors. Every unsupported or partial capability has an elegant state that explains why and suggests another in-platform path.

See Fahd on your own course.

We'll show you the agent loop and result cards inside your platform.